Anthropic’s Fable 5 is not winning over customers. Given the massive investments in this frontier LLM, there is reason to question the LLM specialist’s business model.
Anthropic is struggling to convince many customers with Fable 5. The company’s most powerful LLM attracts about eleven percent of users. Other cheaper models, such as Opus 4.8 and Opus 5, are managing to appeal to more and more customers. This is according to the Ramp AI Index, which tracks figures on the use of AI models in the US.
The trend is striking, as the so-called frontier models of American AI specialists are, after all, their differentiating capability. Anthropic, as well as OpenAI and Google, invest billions in training these highly advanced and large LLMs. The limited adoption of Fable 5 shows that AI development is reaching a point where more powerful does not necessarily mean better. This, in turn, makes it difficult to continue justifying the massive investments in training.
Better but also more expensive
The cause is not hard to find. Fable 5 is not only Anthropic’s best LLM but also its most expensive. In contrast, the company’s Opus models are already very good. It seems that customers are finding that Opus does what it needs to do today. The continued popularity of the smaller Sonnet models supports this line of reasoning: customers are increasingly considering which model is most suitable for their agent or prompt.
The situation at Anthropic thus reflects the broader market situation. There, the major American AI builders are facing increasing competition from smaller, but more efficient and, above all, freely available LLMs, often trained in China. Based on this trend, the number of users who need an LLM like Fable 5 and are willing to spend the budget for it does not seem infinite.
Growth with a caveat
This is complex news for Anthropic, which still wants to be valued at at least one trillion in a future IPO. The company is growing rapidly, however, with a sevenfold increase in revenue since the beginning of this year.
After a first quarter with operational profit, Anthropic would also have to repeat that feat in the next quarter. It is unclear to what extent the costs for training the next frontier models are effectively fully covered in that quarterly calculation, but it does show that Anthropic is no longer the massive money pit of the past.
Chinese threat
Nevertheless, the limited adoption of Fable 5 is a cause for concern. After all, it already shows that there could be a limit to the company’s growth. Furthermore, in the market for frontier models, Anthropic has little sustainable competition, except for Google and OpenAI. This is different for the slightly older LLMs, where the company must accept narrower margins to remain competitive against, for example, DeepSeek or Z-AI.
In China, the ambition to create high-quality AI independently continues to grow. Cloud giant Alibaba will invest the equivalent of 8.8 billion dollars in AI capabilities and infrastructure. That company’s latest LLM (Qwen 3.8-Max), like American models, is said to be well-capable of writing and adapting complex code without too much human oversight.
